swift-code-review — Claude Code skill: Swift review
swift-code-review is a Claude Code skill that checks Swift code for concurrency, error-handling, and memory issues. It activates for .swift reviews involving async/await, actors, or Sendable.
Apache-2.0 253 installs ★ 89 updated 2026-09-15
npx skills add existential-birds/beagle --skill swift-code-review -a claude-codeWhat swift-code-review does
The skill checks changed code for hazards such as force unwraps on runtime values, retain cycles, empty error handlers, and tasks that ignore cancellation. It reads the full enclosing type or member for each changed file, then records the Swift language mode and the checklist areas relevant to the diff.
Findings include a file and line, a severity label, and an explanation. Separate reference files cover concurrency, observation, error handling, and common mistakes. Before reporting an issue, the skill requires the review-verification-protocol at ../review-verification-protocol/SKILL.md; that file must be available for the review workflow to complete.
What it helps with
- spot unsafe
Sendableand@unchecked Sendableusage - check cancellation in long-running tasks
- find retain cycles involving closures or delegates
- review error handling and force unwraps
- report findings with line references and severity
Example prompts for the agent
- Review the changed Swift files for concurrency issues and cite each finding.
- Check this code for retain cycles and error-handling problems.
- Review the changes in
Package.swiftand select the relevant checklist areas.
How to install swift-code-review
1Skills CLI (Vercel)
npx skills add existential-birds/beagle --skill swift-code-review -a claude-code2Manually: put the files in place
~/.claude/skills/swift-code-review/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/swift-code-review/
1Skills CLI (Vercel)
npx skills add existential-birds/beagle --skill swift-code-review -a codex2Manually: put the files in place
.agents/skills/swift-code-review/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add existential-birds/beagle --skill swift-code-review -a cursor2Manually: put the files in place
.cursor/skills/swift-code-review/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install existential-birds/beagle plugins/beagle-ios/skills/swift-code-review2Manually: put the files in place
.github/skills/swift-code-review/SKILL.md1Skills CLI (Vercel)
npx skills add existential-birds/beagle --skill swift-code-review -a gemini-cli2Platform CLI
gemini skills install https://github.com/existential-birds/beagle --path plugins/beagle-ios/skills/swift-code-reviewManually: put the files in place
.devin/skills/swift-code-review/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add existential-birds/beagle --skill swift-code-review -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about swift-code-review
How do I install swift-code-review in Claude Code?
Run npx skills add existential-birds/beagle --skill swift-code-review -a claude-code. The skill file is at ~/.claude/skills/swift-code-review/SKILL.md.
What does swift-code-review do?
It reviews Swift code for concurrency, memory-management, and error-handling issues, then formats findings with a file, line, and severity.
What does swift-code-review check before reporting a finding?
It establishes the Swift language mode, reads the enclosing changed symbol, and selects relevant checklist areas. Each finding must also pass the review-verification-protocol.
Source text
SKILL.md · Apache-2.0
--- name: swift-code-review description: Reviews Swift code for concurrency safety, error handling, memory management, and common mistakes. Use when reviewing .swift files for async/await patterns, actor isolation, Sendable conformance, or general Swift best practices. --- # Swift Code Review ## Review Workflow Follow this sequence **in order**. Do not emit findings until every **Pass** below is satisfied. 1. **Swift / toolchain baseline** — Establish language and tooling context: `Package.swift` `// swift-tools-version` and any per-target Swift language version or `swiftSettings` in the manifest; for Xcode, `SWIFT_VERSION` (or equivalent) in project or target build settings; note if review is single-file only. **Pass:** You state a concrete Swift language version or mode (e.g. Swift 6 language mode, tools 5.10) *before* advice that depends on strict concurrency, migration-only syntax, or SDK availability. 2. **Read surrounding code** — For each changed `.swift` file, read the full enclosing type, function, method, or property that contains the edits, not only the diff hunk. **Pass:** At least one full enclosing symbol (type or member) containing the change was read per changed file. 3. **Scope the checklist** — Using [Quick Reference](#quick-reference), decide which [Review Checklist](#review-checklist) rows and [references](#when-to-load-references) apply; open those reference files; skip rows clearly unrelated to the diff. **Pass:** The review (or working notes) lists which checklist areas you applied, or marks areas N/A with a one-line reason tied to the diff (e.g. “no SwiftUI / @Observable in change”). 4. **Pre-report verification** — Load and follow [review-verification-protocol](../review-verification-protocol/SKILL.md). **Pass:** That skill’s **Hard gates (sequenced)** are satisfied for each finding you will report (full symbol read, usage search before “unused”, caller checked before “missing handling”, severity calibrated, `[FILE:LINE]` proof). ## Hard gates (same sequence, shorter) | Step | Objective pass condition | | --- | --- | | 1 | Swift version/mode (or explicit single-file limitation) recorded before version- or SDK-gated advice. | | 2 | Full enclosing symbol read per changed file, not diff-only. | | 3 | Checklist areas + references listed or N/A with diff-tied reason. | | 4 | `review-verification-protocol` completed for every reported issue. | ## Output format Report findings as: ```text [FILE:LINE] ISSUE_TITLE Severity: Critical | Major | Minor | Informational Description of the issue and why it matters. ``` ## Quick Reference | Issue Type | Reference | |------------|-----------| | async/await, actors, Sendable, Task | [references/concurrency.md](references/concurrency.md) | | @Observable, @ObservationIgnored, @Bindable | [references/observable.md](references/observable.md) | | throws, Result, try?, typed throws | [references/error-handling.md](references/error-handling.md) | | Force unwraps, retain cycles, naming | [references/common-mistakes.md](references/common-mistakes.md) | ## Review Checklist - [ ] No force unwraps (`!`) on runtime data (network, user input, files) - [ ] Closures stored as properties use `[weak self]` - [ ] Delegate properties are `weak` - [ ] Independent async operations use `async let` or `TaskGroup` - [ ] Long-running Tasks check `Task.isCancelled` - [ ] Actors have mutable state to protect (no stateless actors) - [ ] Sendable types are truly thread-safe (beware `@unchecked`) - [ ] Errors handled explicitly (no empty catch blocks) - [ ] Custom errors conform to `LocalizedError` with descriptive messages - [ ] Nested @Observable objects are also marked @Observable - [ ] @Bindable used for two-way bindings to Observable objects ## When to Load References - Reviewing async/await, actors, or TaskGroups → concurrency.md - Reviewing @Observable or SwiftUI state → observable.md - Reviewing error handling or throws → error-handling.md - General Swift review → common-mistakes.md ## Review Questions 1. Are async operations that could run concurrently using `async let`? 2. Could actor state change across suspension points (reentrancy bug)? 3. Is `@unchecked Sendable` backed by actual synchronization? 4. Are errors logged and presented with helpful context? 5. Could any closure or delegate create a retain cycle?
© existential-birds · Apache-2.0 · existential-birds/beagle